The Residual Approach of Variance Decomposition for Stock Returns: A Critique

Abstract

This study criticizes the variance decomposition of stock returns yielded by the residual approach. In the residual approach, the dividend growth component is obtained as the residual after the future return component is estimated. This residual approach has two drawbacks. First, it is prone to understate the contribution of dividend growth component if dividend growth rates and discount rates move together. The second problem arises when stock price has a nonlinear relation with dividend. This paper compares the residual approach with the direct approach in which the dividend growth component is estimated directly from the dynamics of dividend. The results are as follows: In the direct approach, the variability in stock return is explained mostly by the dividend growth component. In the residual approach, however, a fraction of future returns can be forecast by the dividend-price ratio, and sometimes a large contribution is assigned to the future return component. When we perform a variance decomposition using the simulation method, we find that the residual approach has a tendency to overstate the contribution of future return component and understate the contribution of dividend growth component as compared to the direct approach.